6 ms·

For some reason, I've never been able to get the knack of regular expressions. This is the best introduction to them that I've seen, though I wish I'd read it y

For some reason, I've never been able to get the knack of regular expressions. This is the best introduction to them that I've seen, though I wish I'd read it years ago. However, if you've tinkered with regex at all ever, which I suspect virtually everyone here has, you'll know this. (I know it, and I hate and avoid regular expressions and am not even a programmer.)

malkia 16y ago
There a couple of good programs for testing regex-es live. My favourite is Regex Coach by the awesome Edi Weitz: http://weitz.de/regex-coach/ http://weitz.de/regex-coach/ - It's Windows only though There are many others, but his one thought me quite a lot
- gnubardt 16y ago
  vim can be a powerful tool for visualizing regular expressions as well. paste in the text and set incsearch and set hlsearch to match on partial expressions and highlight matches. this allows you to distinctly see the effects of a regex as you type it. also, python's regex module can show you how it interprets a regular expression by passing the re.DEBUG flag to re.compile: >>> import re >>> re.compile('a*b+', re.DEBUG) max_repeat 0 65535 literal 97 max_repeat 1 65535 literal 98 <_sre.SRE_Pattern object at 0x100460230>
Natsu 16y ago
It's weird, because they've always just clicked for me. I mean, you have text and you compare it to a pattern. The regex is just a crazy shorthand for the pattern you're looking for. What pattern should you match? Well, first write down examples of what you want to match and see what they have in common. I think it's easier to get if you match a pattern to some text by hand, taking a simple pattern like ^a+bc$ and trying to see if it matches: Here are a few test cases: "aaaaaaaaabbbc" ? 1) ^a+ means it starts with 1 or more 'a's, so those eat all my 'a's and I'm matching bbbcc against 'bc' now 2) b* means 0 or more 'b's; those eat up everything but the c at the end. 3) c$ matches exactly one c at the end, so yeah, we have a match. If there had been one more c (or anything else), the match would fail because it wouldn't hit the $ (which is end of string). aaac ? 1) ^a+ eats all the 'a's again. 2) b* ... what b? It's "zero or more" so yes, we can skip it. 3) c$ Yes, there's one c at the end. We have a match. bc ? 1) We need one* or more 'a's at the start. That's the difference between + and . Match fails immediately. acbbc ? 1) ^a+ matches the a. 2) b means it's optional, so we don't fail... yet. But we don't match anything, either. 3) c$ That c matches, but we're not at the end of the string yet. There's still "bbc" left to match and we're out of pattern. Fail. You can do something similar with the tools other people have linked in comments. This was just to give an example pattern and a few test cases to get you started. It's weird, though. Even though this clicks for me, I don't feel like I get monads at all. Yeah, I've read the 3 monad laws (great... so something being a monad means that I can monad it, un-monad it and do something vaguely similar to composition on monads) and I can do some very simple Lisp which uses built-in stuff they say is built from monads (like printing). But I don't know how they work internally at all, or how they're fundamentally different from all the other Lisp statements ("causes side affects" isn't terribly specific... why is printing to the screen a side effect? Is setq a monad?). In other words, there's probably some topic for every programmer that they really have to struggle with. As easily as regexes come to me, I struggle to understand other stuff. I've been at it for quite a while, incidentally. But I'm not going to give up any time soon. So keep working on it.
- lincolnq 16y ago
  I wrote some stuff about monads on HN a while ago that some people found helpful: http://news.ycombinator.com/item?id=247446 http://news.ycombinator.com/item?id=247446
  - Natsu 16y ago
    Thanks. I just read it. It does help, but I think there's still a lot I don't know. I mean, 'bind' and 'return' still seem like the names are backwards to me. Somehow, I feel like 'return' should give me a value, rather than shoving a value into a monad. But that's probably because I'm used to C and its relatives. Perhaps I simply need to read up more on FP itself. I have a huge bookshelf full of half-read books (and yes, plenty of fully-read books) but nowhere near enough free time...
    - lincolnq 16y ago
      Oh. Yeah. The names are poorly chosen, but we're stuck with them. :(
- Natsu 16y ago
  HN ate some stars in the above post and I didn't notice until now. Sorry about that. Look for the italics to see where the stars should be.
Tichy 16y ago
I suspect your eyes tend to glaze over because of the many strange symbols? Just relax - they are not actually hard, it looks more complicated than it is. That there used to be a whole O'Reilly book dedicated to them also made them look more complicated. But there are books for everything (probably even Notepad).

malkia 16y ago

There a couple of good programs for testing regex-es live. My favourite is Regex Coach by the awesome Edi Weitz: http://weitz.de/regex-coach/ http://weitz.de/regex-coach/ - It's Windows only though There are many others, but his one thought me quite a lot

gnubardt 16y ago
vim can be a powerful tool for visualizing regular expressions as well. paste in the text and set incsearch and set hlsearch to match on partial expressions and highlight matches. this allows you to distinctly see the effects of a regex as you type it. also, python's regex module can show you how it interprets a regular expression by passing the re.DEBUG flag to re.compile: >>> import re >>> re.compile('a*b+', re.DEBUG) max_repeat 0 65535 literal 97 max_repeat 1 65535 literal 98 <_sre.SRE_Pattern object at 0x100460230>

Natsu 16y ago

It's weird, because they've always just clicked for me. I mean, you have text and you compare it to a pattern. The regex is just a crazy shorthand for the pattern you're looking for. What pattern should you match? Well, first write down examples of what you want to match and see what they have in common. I think it's easier to get if you match a pattern to some text by hand, taking a simple pattern like ^a+bc$ and trying to see if it matches: Here are a few test cases: "aaaaaaaaabbbc" ? 1) ^a+ means it starts with 1 or more 'a's, so those eat all my 'a's and I'm matching bbbcc against 'bc' now 2) b* means 0 or more 'b's; those eat up everything but the c at the end. 3) c$ matches exactly one c at the end, so yeah, we have a match. If there had been one more c (or anything else), the match would fail because it wouldn't hit the $ (which is end of string). aaac ? 1) ^a+ eats all the 'a's again. 2) b* ... what b? It's "zero or more" so yes, we can skip it. 3) c$ Yes, there's one c at the end. We have a match. bc ? 1) We need one* or more 'a's at the start. That's the difference between + and . Match fails immediately. acbbc ? 1) ^a+ matches the a. 2) b means it's optional, so we don't fail... yet. But we don't match anything, either. 3) c$ That c matches, but we're not at the end of the string yet. There's still "bbc" left to match and we're out of pattern. Fail. You can do something similar with the tools other people have linked in comments. This was just to give an example pattern and a few test cases to get you started. It's weird, though. Even though this clicks for me, I don't feel like I get monads at all. Yeah, I've read the 3 monad laws (great... so something being a monad means that I can monad it, un-monad it and do something vaguely similar to composition on monads) and I can do some very simple Lisp which uses built-in stuff they say is built from monads (like printing). But I don't know how they work internally at all, or how they're fundamentally different from all the other Lisp statements ("causes side affects" isn't terribly specific... why is printing to the screen a side effect? Is setq a monad?). In other words, there's probably some topic for every programmer that they really have to struggle with. As easily as regexes come to me, I struggle to understand other stuff. I've been at it for quite a while, incidentally. But I'm not going to give up any time soon. So keep working on it.

lincolnq 16y ago
I wrote some stuff about monads on HN a while ago that some people found helpful: http://news.ycombinator.com/item?id=247446 http://news.ycombinator.com/item?id=247446
- Natsu 16y ago
  Thanks. I just read it. It does help, but I think there's still a lot I don't know. I mean, 'bind' and 'return' still seem like the names are backwards to me. Somehow, I feel like 'return' should give me a value, rather than shoving a value into a monad. But that's probably because I'm used to C and its relatives. Perhaps I simply need to read up more on FP itself. I have a huge bookshelf full of half-read books (and yes, plenty of fully-read books) but nowhere near enough free time...
  - lincolnq 16y ago
    Oh. Yeah. The names are poorly chosen, but we're stuck with them. :(
Natsu 16y ago
HN ate some stars in the above post and I didn't notice until now. Sorry about that. Look for the italics to see where the stars should be.

Tichy 16y ago

I suspect your eyes tend to glaze over because of the many strange symbols? Just relax - they are not actually hard, it looks more complicated than it is. That there used to be a whole O'Reilly book dedicated to them also made them look more complicated. But there are books for everything (probably even Notepad).